The TAKE IT DOWN Act, signed into law on May 19, 2025, marked a significant milestone in addressing the harms caused by AI-generated content. The act was galvanized by stories like that of Elliston Berry, a 14-year-old who was victimized by AI-generated nude images created by a classmate. This bipartisan legislation criminalizes the knowing publication of non-consensual intimate imagery (NCII), including deepfakes and digital manipulations. The act passed the House of Representatives with an overwhelming vote of 409-2 and was supported by both parties due to its focus on privacy and consent for minors. First Lady Melania Trump also publicly endorsed the bill, highlighting the widespread concern over the impact of social media on young people.
The TAKE IT DOWN Act introduces criminal penalties for the publication of NCII and requires platforms to set up a process to remove such content within 48 hours of a valid written notification from a victim. The Federal Trade Commission (FTC) will enforce these takedown processes, which platforms must implement by May 19, 2026. Violations of the act can result in fines or imprisonment for up to two years, with increased penalties if the content depicts minors. Despite concerns from some organizations about potential overbroad censorship, the mechanisms for protecting children online were widely praised.
While the TAKE IT DOWN Act is a crucial step, it is just the beginning of congressional action needed to mitigate harms related to AI. In the 118th Congress, over 150 AI-related bills were introduced, but none garnered the same level of support as the TAKE IT DOWN Act. Some notable measures included the Algorithmic Accountability Act, which aims to require companies using automated decision-making systems to conduct impact assessments for bias, fairness, privacy, and security. Another proposal, the National AI Commission Act, sought to establish a bipartisan commission to review the U.S. approach to AI regulation and recommend regulatory structures. Other bills addressed algorithmic discrimination, harms to civil rights, information integrity, elections, and personality or likeness rights.
The TAKE IT DOWN Act’s swift passage can be attributed to its focus on a highly visible, emotionally salient harm that was easily explainable to the public. By July 2025, at least 47 states had enacted laws regulating deepfakes, with many of these laws providing victims of NCII with recourse similar to the TAKE IT DOWN Act. These state laws often emphasize children’s safety and preventing irreparable damage, aligning with other bills gaining attention in Congress, such as the Kids Online Safety Act (KOSA). Both KOSA and the TAKE IT DOWN Act raise concerns about potential censorship and the intersection with Section 230, which provides platforms immunity for user-generated content.
Algorithmic discrimination, a persistent and well-documented harm, impacts various aspects of life, including housing, credit, policing, and hiring decisions. Vulnerable populations are disproportionately affected due to the bias in the data fed into these systems. For example, housing algorithms can mistakenly deny applicants based on criminal histories or identify applicants with someone else’s record. Similarly, credit scoring and eviction histories, which reflect historical discrimination, can deny people of color loans or issue them higher rates. In the hiring process, resume screening technology can discriminate against applicants due to race, gender, and the intersection of these identities. However, proving algorithmic discrimination is challenging and time-consuming, requiring a longitudinal compilation of evidence and persuading regulators and lawmakers to act.
The release of ChatGPT in 2022 brought generative AI into the public eye, allowing users to witness the evolution of image-generation systems in real time. Despite this increased visibility, people subjected to algorithmic discrimination still lack transparency regarding how these models impact their lives. Congress must address all AI harms, not just the more visible ones. AI models are often described as “opaque” or “black boxes,” but this does not mean policymakers can ignore the risks they pose. Stronger governance and enforcement systems are needed to identify and respond to these quieter risks effectively. Requiring companies to conduct ongoing impact assessments of automated decision systems used in critical areas would help test for material negative impacts and document how such harms are mitigated.
Protecting Americans from AI harms requires more than safeguards against the most visible problems. It involves bolstering data access, auditing regimes, enforcement capacity, transparency, and considering more robust harms in legal frameworks. Legislative attention is also needed for stronger whistleblower protections. Employees and contractors are often the first to identify algorithmic harms but may lack the legal protections to report them or face company retaliation. Strengthening whistleblower channels, especially around trade secret and retaliation protections, could increase transparency on harms that would otherwise be difficult to spot.
At the state level, attorneys general in several jurisdictions have issued guidance on how emerging AI technologies can perpetuate algorithmic discrimination and violate existing antidiscrimination laws. State legislators have also been active, with Colorado and Illinois enacting laws with protections against AI-facilitated discrimination. Legislative proposals have been floated in California, Connecticut, Virginia, and Texas, among other states. Many of these states are influenced by the EU AI Act, which requires impact assessments and auditing for AI systems to evaluate potential discriminatory harms.
Limitaciones potenciales
Sin embargo, recientes amenazas legislativas y ejecutivas para prevenir los esfuerzos estatales para regular el AI pueden limitar el alcance de estas acciones. Es importante tener en cuenta que más robustas protecciones pueden proporcionar transparencia y agencia adicionales en la cara de tal discriminación.
